AWS expands model choice in Bedrock while AgentCore cuts cold starts and Strands trims token use by 28 percent.
AWS shipped September updates across Amazon Bedrock, AgentCore, and Strands. Amazon Bedrock Managed Agents, powered by OpenAI models, entered public preview, and the AgentCore runtime now starts serverless agents faster with lower cold start latency and pay-as-you-go pricing.
AWS frames model choice, runtime, and tooling as the three layers where enterprises need options. The company says customers now weigh cost against benefit per use case rather than picking models on raw performance, especially when delegating work to agents in production systems.
Builders gain a way to run OpenAI models on AWS without moving data outside the account, keeping existing IAM roles and CloudTrail audit trails. Durable sessions and built-in human approval workflows cut development overhead and reduce risk for teams putting agents into regulated workflows.
Strands released an open source agent harness that matches popular harnesses on accuracy while using 28 percent fewer tokens, and it runs a production-ready agent from one line of Python or TypeScript. Watch how pay-as-you-go runtime pricing and session scale-to-zero shift agent cost planning.
What matters
- Amazon Bedrock Managed Agents, powered by OpenAI models, entered public preview in September.
- Teams keep data inside AWS, reuse IAM permissions, and audit activity through CloudTrail.
- Watch AgentCore runtime adoption as cold start latency and idle scaling reshape agent cost models.
Why it matters
Watch AgentCore runtime adoption as cold start latency and idle scaling reshape agent cost models.
This GenAI News article was prepared in original wording using reporting and materials published by AWS Machine Learning Blog. Source reference: https://aws.amazon.com/blogs/machine-learning/icymi-what-landed-for-ai-builders-in-september-2026/.
Drafted by the GenAI News review pipeline.